~10% hallucinated sentences: guardrails for lifelogging with LLMs
thejaan · x · 2026-09-24
In a follow-up to his lifelogging thread, Jaan Altosaar details that about 10% of the LLM-generated sentences are made up (e.g. one claimed his dad was recovering from surgery in Estonia), so he listens with vigilance. His guardrails: no memory across sessions, no follow-up questions to avoid reassurance-seeking, a fresh session every 5-10 weeks, and treating it as no substitute for therapy.
More from Apps
- ElevenLabs voice generation comes to Meta's personal AI agent Muse — lukeharries · 2026-09-24
- ElevenLabs' Big Week: AI Receptionist, Studio 4.0, and Scribe v2 Medical Cut Errors 35% — lukeharries · 2026-09-24
- A heavy ChatGPT user can't find the job that belongs to OpenAI's Muse — chufan_shi · 2026-09-24
- Grok bot gets 1Password integration with dedicated vaults for selective key sharing — altryne · 2026-09-24
- Merge launches Universal Context Layer, a shared company brain for every AI agent — shensi · 2026-09-24
- Delegation is hard: why in-app agents may beat general-purpose agents for consumers — nathanborror · 2026-09-24